A markov chain is a process that describes a sequence of events, where the probability of a
specific event depends solely on the state attained at the previous event. For a more complete
description of these, feel free to consult the Wikipedia article on markov chains. Concretely, markov chains are used for autocompletion when typing. You can see
this in action by writing a word on your phone, and if you have the setting enabled, you should see
some suggestions for the next word above your keyboard. I decided to implement this in a more naive
way, because the program can only get its data from a file containing text that it will learn from,
which is parsed to make what I call a "model". This model is then stored as a JSON file (although
it could technically be stored in a much more space-efficient format), so that it can be used later
on without having to repeatedly parse the input file. A model basically states, for each word present
in the input file, the probability of a certain word appearing right after it. The big.json model in the GitHub repository is created from a few books, taken from Project Gutenberg. In this model, the word "zinc" has a 50% chance to be followed by "and", 25% chance to be
followed by "Immersion" and a 25% chance to be followed by "A". This is essentially what a
markov chain is. After parsing the JSON file and storing the information into a data structure
that is optimized for our needs, the program will spout whole paragraphs on autocomplete from
one starting word, even if they tend to be nonsensical at times, which is normal, since the
words proposed by the model are just suggestions of the most common words.
Example prompt using a model based on the Frankenstein book:
Starting word: Why
did not be at the same time I had been out of my eyes of my love and the
sickening oppression of the first saw a few words only took place of the same time I shall be a few
words of the same time I had been so miserable of her guilt Example prompt using a model based on the complete works of Shakespeare:
Starting word: Thus
may be the law and never gave to the King hath pleas’d to her my lord of your Grace to the King Richard
and the world but the rest at the eye and my lord you come to the world I will not a man that to
be not in the body of the world will I will I will be so much of the world to beat aside the King
Richard and the other Diomed faith in one that I will you sir if thou that you as one that which
is an excellent good lord Safe mayst be no more advice